Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intelligent Model Update Strategy for Sequential Recommendation

Zheqi Lv, Wenqiao Zhang, Zhengyu Chen, Shengyu Zhang, Kun Kuang

TL;DR

The paper addresses the inefficiency of frequent cloud-to-edge parameter updates in edge-cloud collaborative sequential recommendations. It introduces IntellectReq, a lightweight edge-side controller comprising MRD and DM that decides when cloud-generated dynamic parameters are truly beneficial, and maps real-time user behavior to a Gaussian latent space to estimate uncertainty. Building on EC-CDR, IntellectReq enables revenue-aware, budget-guided edge updates by predicting mis-recommendations and modeling data-shift uncertainty, with an MRD-driven threshold governing updates. Empirical evaluation on CDs, Electronic, and Douban Books benchmarks shows that IntellectReq delivers competitive accuracy with reduced communication and computing costs, achieving substantial resource savings while maintaining generalization and adaptability in dynamic user environments.

Abstract

Modern online platforms are increasingly employing recommendation systems to address information overload and improve user engagement. There is an evolving paradigm in this research field that recommendation network learning occurs both on the cloud and on edges with knowledge transfer in between (i.e., edge-cloud collaboration). Recent works push this field further by enabling edge-specific context-aware adaptivity, where model parameters are updated in real-time based on incoming on-edge data. However, we argue that frequent data exchanges between the cloud and edges often lead to inefficiency and waste of communication/computation resources, as considerable parameter updates might be redundant. To investigate this problem, we introduce Intelligent Edge-Cloud Parameter Request Model, abbreviated as IntellectReq. IntellectReq is designed to operate on edge, evaluating the cost-benefit landscape of parameter requests with minimal computation and communication overhead. We formulate this as a novel learning task, aimed at the detection of out-of-distribution data, thereby fine-tuning adaptive communication strategies. Further, we employ statistical mapping techniques to convert real-time user behavior into a normal distribution, thereby employing multi-sample outputs to quantify the model's uncertainty and thus its generalization capabilities. Rigorous empirical validation on four widely-adopted benchmarks evaluates our approach, evidencing a marked improvement in the efficiency and generalizability of edge-cloud collaborative and dynamic recommendation systems.

Intelligent Model Update Strategy for Sequential Recommendation

TL;DR

The paper addresses the inefficiency of frequent cloud-to-edge parameter updates in edge-cloud collaborative sequential recommendations. It introduces IntellectReq, a lightweight edge-side controller comprising MRD and DM that decides when cloud-generated dynamic parameters are truly beneficial, and maps real-time user behavior to a Gaussian latent space to estimate uncertainty. Building on EC-CDR, IntellectReq enables revenue-aware, budget-guided edge updates by predicting mis-recommendations and modeling data-shift uncertainty, with an MRD-driven threshold governing updates. Empirical evaluation on CDs, Electronic, and Douban Books benchmarks shows that IntellectReq delivers competitive accuracy with reduced communication and computing costs, achieving substantial resource savings while maintaining generalization and adaptability in dynamic user environments.

Abstract

Modern online platforms are increasingly employing recommendation systems to address information overload and improve user engagement. There is an evolving paradigm in this research field that recommendation network learning occurs both on the cloud and on edges with knowledge transfer in between (i.e., edge-cloud collaboration). Recent works push this field further by enabling edge-specific context-aware adaptivity, where model parameters are updated in real-time based on incoming on-edge data. However, we argue that frequent data exchanges between the cloud and edges often lead to inefficiency and waste of communication/computation resources, as considerable parameter updates might be redundant. To investigate this problem, we introduce Intelligent Edge-Cloud Parameter Request Model, abbreviated as IntellectReq. IntellectReq is designed to operate on edge, evaluating the cost-benefit landscape of parameter requests with minimal computation and communication overhead. We formulate this as a novel learning task, aimed at the detection of out-of-distribution data, thereby fine-tuning adaptive communication strategies. Further, we employ statistical mapping techniques to convert real-time user behavior into a normal distribution, thereby employing multi-sample outputs to quantify the model's uncertainty and thus its generalization capabilities. Rigorous empirical validation on four widely-adopted benchmarks evaluates our approach, evidencing a marked improvement in the efficiency and generalizability of edge-cloud collaborative and dynamic recommendation systems.
Paper Structure (28 sections, 26 equations, 11 figures, 5 tables)

This paper contains 28 sections, 26 equations, 11 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Methodology
  4. Problem Formulation
  5. IntellectReq
  6. The framework of EC-CDR
  7. Mis-Recommendation Detector
  8. Distribution Mapper
  9. On-edge Model Update
  10. Experiments
  11. Experimental Setup.
  12. Experimental Results.
  13. Quantitative Results.
  14. Mis-recommendation score and profit.
  15. Ablation Study.
  16. ...and 13 more sections

Figures (11)

  • Figure 1: (a) outlines the evolution of recommendation systems from On-Edge Static Model to Edge-Cloud Collaborative and Dynamic Recommendation (EC-CDR). (b) overviews our IntellectReq. (c) compares the three methods and our IntellectReq (Communication Frequency 10% (IntellectReq) $\ll$ 100% (EC-CDR), AUC: 0.8562 (IntellectReq) $\approx$ 0.8581 (EC-CDR) or Communication Frequency 3% (IntellectReq) $\ll$ 100% (EC-CDR), HR@20: 0.6478 (IntellectReq) $\approx$ 0.6478 (EC-CDR))
  • Figure 2: Domain numbers of users.
  • Figure 3: Overview of the IntellectReq. (a) describes the conventional recommendation model. (b) describes the EC-CDR. (c) and (d) respectively showcase the IntellectReq, and its Mis-Recommendation Detector and Distribution Mapper modules.
  • Figure 4: Overview of the proposed Distribution Mapper. Training procedure: The architecture includes Recommendation Network, Prior Network, Posterior network and Next-item Perdition Network. Loss consists of the classification loss and the KL-Divergence loss. Inference procedure: The architecture includes Recommendation Network, Prior Network and Next-item Perdition Network. The uncertainty is calculated by the multi-sampling output.
  • Figure 5: Performance w.r.t. Request Frequency curve based on previous 1 time difference on-edge dynamic model.
  • ...and 6 more figures